1. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 1 ECSE-6660: Broadband Networks Homework 2 Please Submit Online in the WebCT dropbox Deadline : 19 th Feb (non-tape-delayed) Feb 26th (tape-delayed) Homework 2 Please Submit Online in the WebCT dropbox Deadline : 19 th Feb (non-tape-delayed) Feb 26th (tape-delayed)

2. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 2 I. Reading Assignment & Quick Questions (100%) Reading assignments count for a substantial part of homework credit Carefully review slide sets 3,4,5; Read Chapter 4, 15 of S. Keshav’s book, and Chap 6, 13, Secs: 10.1/10.2 of Ramaswami/Sivarajan. Then answer the following quick true/false questions that test your knowledge. Please submit the electronic version of this powerpoint file with your answers. (Cut-and-paste the tick (  ) over the appropriate boxes on the left) [60 questions; 5/3 points per question] T F  ”Switched” services offer the capability to “dial-up” a data service between two points  Switched services usually offer distance-sensitive pricing plans just like T-carrier services  Signaling and path-pinning is an important commonality between the “virtual-circuit” and “circuit-switched” network architectures  The data-plane “labels” used in virtual circuits have a global meaning, I.e. they represent identifiers that can be interpreted in the same way by any node in the network

3. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 3  QoS capabilities are mandatory in virtual-circuit based networks  X.25 uses a connectionless packet-switching approach and does not nail down paths before transmitting data  X.25 is an example of a heavyweight protocol design where similar functions are found in multiple layers.  The internet architecture tends to “couple” a variety of protocol design issues (eg: combines QoS with routing with signaling etc)  The difference between “connection-oriented” (eg: TCP) and “virtual circuit” (eg: frame relay or ATM) method is that the former only establishes an end-to-end state association, whereas the latter establishes state at nodes along the path.  X.25’s link layer (LAP-B) is derived from HDLC  X.25 uses LAP-B and ISDN uses LAP-D, both of which are derived from HDLC  LAP-B uses the balanced configuration mode of HDLC where both nodes can act as a transmitter or receiver  X.25 “sets-up” a link before transmitting packets similar to how telephony would “setup” an end-to-end circuit before transmitting  Frame relay dramatically simplifies the link layer equivalent of X.25  Frame relay & frame switching implement LAP-F control protocol at all nodes  The PRI service of ISDN offers 128 kbps and a D-control channel  ISDN, like SONET, offers both channelized and concatenated channels

4. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 4  ISDN extends the concept of digital modulation, digital transmission and common channel signaling all the way to the home.  X.25 and frame relay offer a significant cost advantage over leased lines, but offer reduced or no QoS guarantees; and a packet-switching service instead of a circuit-switching service  Frame relay uses one control VC for all its data VCs; X.25 uses in-band control signaling leading to a complex frame format  The CIR parameter of frame relay represents the maximum rate at which a user may transmit; all packets above this rate are dropped  A user may transmit an unbounded number of simultaneous frames at the rate PIR.  The DE bit if set does not guarantee that the packet will be dropped, but assigns it a higher effective drop probability during congestion  The frame-relay receiver responds to FECN bit settings by reducing its transmission rate  In BECN, an intermediate relay device could immediately inform the source about congestion rather than mark a bit on a forward-going packet.  ATM networks were designed to offer the best of the two worlds of telephony (eg: reliability, QoS) and data networking (lower cost, flexibility, statistical multiplexing gains)  ATM uses variable sized packets  The 53-byte cell is an efficient way to packetize data  The term UNI refers to the user-to-network interface  PNNI is an example of a network-to-network interface protocol  One of the important functions of AAL is segmentation-and-reassembly (SAR)  ATM’s HEC field serves dual-use as a cell-boundary-synchronization hook and for header error detection

5. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 5  A signaling protocol fundamentally maps global addresses (and path specifications) to local labels (it may optionally reserve resources)  An SVC is an example of an “on-demand” or “dial-up” virtual circuit, whereas a PVC is a more permanent VC.  ATM switches swap incoming labels with outgoing labels on ATM cells (unlike IP routers which do not change the IP address fields on IP packets)  Fixed sized cells tend to reduce average queuing delay in underloaded, bursty systems and also make it easier to build fast switches  The CBR service is most useful for variable-bit-rate, adaptive data transport  The key difference between GFR and UBR is that the former allows the minimum frame rates (instead of minimum cell rates) and emulates an equivalent of a frame-relay VC  ABR provides a feedback-based traffic management feature that is not available in UBR  In a signaled architecture, routing is used to guide the signaling call  PNNI does not support QoS routing  ATM uses a 20-byte global address during the signaling phase and 4-byte labels in the data-plane (I.e. in cells)  Ethernet uses flat addresses whereas IP and ATM use hierarchical addressing  Unlike IP routing ATM also allows peer-group IDs to be encoded within the ATM address  Like IP, ATM uses hop-by-hop routing and avoids the use of source-routing  A DTL is a stack-like mechanism to specify a source-route in PNNI

6. Shivkumar Kalyanaraman Rensselaer Polytechnic Institute 6  A node at the lowest level of the PNNI hierarchy can see only the internal topology of the peer group it is in.  Call admission control (CAC) is needed only if QoS is required  Crank-back is a process where the signaling can backtrack to a prior node if CAC fails at a node  ATM traffic management offers a combination of both open-loop and closed-loop mechanisms  GCRA is implemented using leaky buckets and is an example of a traffic shaping method  ATM inter-networking can happen at the link-layer (LANE), network layer (RFC 1577, MPOA, NHRP)  A fundamental problem in IP-over-ATM inter-networking is how to do address resolution (I.e. IP address to ATM address mapping) and the answer is usually a server-based approach  ATM LANE offers virtual LAN capabilities  RFC 1577 specifies packet encapsulation and RFC 1483 specifies address resolution methods  The problem with RFC 1577 LIS structure is that two hosts on the same ATM network, but different LIS’es will need to go through (a potentially slow) router to communicate  NHRP allows servers to relay address-resolution across LIS’es  MPOA attempts to integrate the benefits of LANE and NHRP  MPLS solves both the data-plane and control-plane IP-over-ATM mapping woes by abandoning the overlay model in favor of a hybrid mapping model  A key problem in traffic engineering is to define traffic aggregates and then map them to an explicitly setup path